In the past three years, the narrative in the AI server market has been remarkably singular: the entity with the most GPUs holds the key to achieving Artificial General Intelligence (AGI). However, as AI applications evolve from simple chatbots to "Agentic AI" capable of autonomously executing complex workflows, the fundamental computing power requirements within data centers are undergoing a radical transformation. According to market intelligence, HSBC Securities states in a newly released in-depth industry report that an era of "computing rebalancing" has commenced, with the CPU re-emerging as the decision-making hub of AI data centers.
**From Conversation to Action: The Computing Shift Driven by Agentic AI** Early AI models primarily relied on massive GPU clusters for parallel computing, an approach often termed "brute force." In contrast, the workflow of Agentic AI is entirely different; it functions as a "super-orchestrator." The HSBC report highlights that Agentic AI not only generates content but also completes tasks by utilizing Retrieval-Augmented Generation (RAG) and accessing external APIs (such as email systems, web search, and collaborative software). These operations involve significant logical branching, task synchronization, and multi-threaded processing. Comparative data indicates that in Agentic AI tasks, the parallel processing advantage of GPUs is diminished, while the flexibility of CPUs in multi-tasking makes them significantly more effective. According to recent research from Cornell University, approximately 44% of the computing power in an Agentic AI workflow relies on CPUs, which is three to four times higher than the CPU utilization in traditional AI workflows. The CPU is no longer merely a system "assistant" but has become an indispensable "brain" within the computing engine.
**Why Are Server Shipment Forecasts Soaring? Supply Falls Far Short of Demand** This structural shift in computing demand is directly translating into robust support for server shipments. HSBC Securities has substantially raised its baseline growth forecast for global server shipments in 2026 to 20% year-over-year (up from a previous 4%), a figure that far exceeds the general market expectation of 10%-15%. More strikingly, HSBC suggests the "latent demand" could be as high as 60%. Why is the market not fully realizing this demand? The primary reason lies in supply chain bottlenecks. The report emphasizes that current supply shortages for CPUs and DRAM memory are as high as 30%-40%, while core components like SSDs and power management chips also face supply gaps of 10%-30%. This implies that global server shipment growth over the next two years will be a constrained, incremental release. This persistent supply shortage has led to a substantial backlog of orders, which not only secures the shipment rationale for 2026-2027 but also leads HSBC to believe the global server growth cycle could extend into 2028.
**Capital Expenditure Surge: The 'Additional Fuel' from Policy** Fueling this hardware boom is not just technological advancement but also policy incentives. According to HSBC's analysis, with the implementation of the relevant US legislation, companies can benefit from a 100% bonus depreciation and immediate R&D expense deductions for purchasing data center assets, including servers. This provides a powerful financial incentive for Cloud Service Providers (CSPs) to invest in hardware. HSBC predicts that the combined capital expenditure of major CSPs will reach $743 billion in 2026, representing an 81% year-over-year increase. The allocation of this capital is undergoing a structural shift: in 2025, approximately 35% of CSP capital expenditure was directed toward servers; by 2027, this proportion is projected to jump to 59%. This reallocation of resources from "land and construction" to "computing equipment" precisely targets benefits for the server supply chain.
**Who Benefits? The 'Rebalancing' Advantage for the Supply Chain** The market focus is shifting from a pure "GPU narrative" to a more balanced "compute node value chain." HSBC points out that as the value contribution of CPUs within server racks increases, server Original Design Manufacturers (ODMs) and key component suppliers are set to become the primary beneficiaries of this computing rebalancing. In summary, the computing arms race has entered a deeper, more complex phase. The past was about "winning with graphics cards," but the future will be about "mastering efficiency through optimal computing resource allocation." The CPU demand resurgence driven by Agentic AI not only injects long-term certainty into the server market but also outlines a clear investment thesis centered on hardware upgrades.
Comments